Glucocorticoid receptor in T cells mediates protectionfrom autoimmunity in pregnancyJan Broder Englera, Nina Kursawea, María Emilia Solanob, Kostas Patasa, Sabine Wehrmanna, Nina Heckmanna,Fred Lühderc, Holger M. Reichardtd, Petra Clara Arckb, Stefan M. Golda,e, and Manuel A. Friesea,1

aInstitut für Neuroimmunologie und Multiple Sklerose, Zentrum für Molekulare Neurobiologie Hamburg, Universitätsklinikum Hamburg-Eppendorf, 20251Hamburg, Germany; bExperimentelle Feto-Maternale Medizin, Klinik für Geburtshilfe und Pränatalmedizin, Universitätsklinikum Hamburg-Eppendorf,20251 Hamburg, Germany; cInstitut für Neuroimmunologie und Institut für Multiple Sklerose Forschung, Universitätsmedizin Göttingen, 37073 Goettingen,Germany; dInstitut für Zelluläre und Molekulare Immunologie, Universitätsmedizin Göttingen, 37075 Goettingen, Germany; and eKlinik für Psychiatrie undPsychotherapie, Campus Benjamin Franklin, Charité Universitätsmedizin, 12203 Berlin, Germany

Edited by Philippa Marrack, Howard Hughes Medical Institute, National Jewish Health, Denver, CO, and approved December 6, 2016 (received for reviewOctober 17, 2016)

Pregnancy is one of the strongest inducers of immunological toler-ance. Disease activity of many autoimmune diseases including mul-tiple sclerosis (MS) is temporarily suppressed by pregnancy, but littleis known about the underlying molecular mechanisms. Here, we in-vestigated the endocrine regulation of conventional and regulatoryT cells (Tregs) during reproduction. In vitro, we found the pregnancyhormone progesterone to robustly increase Treg frequencies viapromiscuous binding to the glucocorticoid receptor (GR) in T cells. Invivo, T-cell–specific GR deletion in pregnant animals undergoing ex-perimental autoimmune encephalomyelitis (EAE), the animal modelof MS, resulted in a reduced Treg increase and a selective loss ofpregnancy-induced protection, whereas reproductive success was un-affected. Our data imply that steroid hormones can shift the immu-nological balance in favor of Tregs via differential engagement of theGR in T cells. This newly defined mechanism confers protection fromautoimmunity during pregnancy and represents a potential target forfuture therapy.

multiple sclerosis | autoimmunity | pregnancy | Treg | steroid hormones

Reproduction is fundamental to the maintenance and evolu-tion of species. To ensure successful pregnancy, mothers

have to establish robust immunological tolerance toward the semi-allogeneic conceptus providing a secure niche for fetal development.Multiple mechanisms have evolved to prevent fetus-directed immuneresponses and alloreactive infiltration of the fetomaternal interface(1). These include creating a privileged local microenvironment thathampers T-cell priming and infiltration (2–4) but also imply globalmodulation of the immune system by pregnancy hormones and theshedding of fetal antigen into the mothers circulation (5).Intriguingly, pregnancy is also well known to suppress the in-

flammatory activity of a number of cell-mediated autoimmune dis-eases, including rheumatoid arthritis (6, 7), autoimmune hepatitis(8), and multiple sclerosis (MS) (9, 10). However, this beneficialeffect is limited to the period of gestation and usually followed by arebound of disease activity postpartum. In the case of MS, thirdtrimester pregnancy leads to a remarkable reduction of the MSrelapse rate (11), which exceeds the effects of most currently avail-able disease-modifying drugs. Similarly, pregnancy as well as treat-ment with pregnancy hormones protect rodents from experimentalautoimmune encephalomyelitis (EAE), a widely used animal modelof MS (12) in both SJL/J and C57BL/6 mice (13–16), underpinningan interaction between pregnancy-related immune and endocrineadaptations and central nervous system (CNS) autoimmunity (17).The sensitive balance between conventional effector T cells

(Tcons) and regulatory T cells (Tregs) has transpired as a commontheme that connects reproductive biology and autoimmunity on amechanistic level (18–21). Tregs are characterized by the tran-scription factor forkhead box P3 (Foxp3) and effectively controleffector responses mounted by Tcons in response to antigen-specificactivation. In the context of pregnancy, Tregs were shown to expandto safeguard against fetal rejection by establishing antigen-directed

immunological tolerance (18, 22). Depletion of Tregs results in fetalloss (23). Pregnancy-related hormonal signals, including estrogensand progesterone (1, 24, 25), appear to support Treg proportionand function, although their exact mode of action remains in-completely understood.In the context of autoimmunity, Tregs are essential in suppressing

autoreactive responses. Foxp3 deficiency results in generalized au-toimmune inflammation evident in scurfy mice (2–4, 26) and inpatients suffering from immunodysregulation polyendocrinopathyenteropathy X-linked syndrome (IPEX) (5, 27). Beyond that,quantitative or functional Treg impairment has been described in anumber of autoimmune diseases, including systemic lupus eryth-ematosus (6, 7, 28, 29), rheumatoid arthritis (8, 30, 31), and type Idiabetes (9, 10, 32). In MS, Tregs were reported to possess di-minished suppressive potential (11, 33, 34) and decreased expres-sion of Foxp3 and immunosuppressive cytotoxic T lymphocyteantigen 4 (CTLA-4) (12, 35–37).However, it is still unknown which pregnancy-related changes

account for the enhanced control of autoreactive responses. Here,we sought to unravel the molecular mechanisms that confer thebeneficial effect of pregnancy on autoimmunity. By studying theimpact of pregnancy and pregnancy hormones on T-cell subsetsincluding Tregs, we provide evidence that differential sensitivity toglucocorticoid receptor (GR) activity is a hitherto unrecognized

Significance

Reproduction in placental mammals relies on potent control ofthe mother’s immune system to not attack the developing fetus.As a bystander effect, pregnancy also potently suppresses theactivity of the autoimmune disease multiple sclerosis (MS). Here,we report that T cells are able to directly sense progesterone viatheir glucocorticoid receptor (GR), resulting in an enrichment ofregulatory T cells (Tregs). By using anMS animal model, we foundthat the presence of the GR in T cells is essential to increase Tregsand confer the protective effect of pregnancy, but not formaintaining the pregnancy itself. Better understanding of thistolerogenic pathway might yield more specific therapeutic meansto steer the immunological balance in transplantation, cancer,and autoimmunity.

Author contributions: J.B.E., S.M.G., and M.A.F. designed research; J.B.E., N.K., M.E.S.,K.P., S.W., and N.H. performed research; F.L., H.M.R., and P.C.A. contributed newreagents/analytic tools; J.B.E., N.K., P.C.A., and M.A.F. analyzed data; and J.B.E. andM.A.F. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. Email: manuel.friese@zmnh.uni-hamburg.de.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1617115114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1617115114 PNAS | Published online January 3, 2017 | E181–E190

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

mechanism shaping the T-cell compartment. By targeted GRdeletion in T cells, we show a selective disruption of pregnancy-induced protection from autoimmunity.

ResultsTreg Dynamics During Pregnancy.We first characterized the relativeabundance and phenotype of Tcons and Tregs throughout preg-nancy and postpartum time points (Fig. 1A). Because the presenceof fetuses that genetically differ from the mother is a prerequisiteof robust induction of immune tolerance (13–15, 18), we performedallogeneic matings of C57BL/6 females with BALB/c males. Weanalyzed cells from paraaortic lymph nodes (LNs), inguinal lymphnodes, and spleen to get insight into local, regional, and systemicchanges, respectively (Fig. 1B). The acquired data indicated thatTreg expansion was most pronounced in the paraaortic lymph nodesthat drain the fetomaternal interface. Frequencies of CD4+Foxp3+

Tregs started to increase from middle to late pregnancy (embryonicdays, E10.5–E18.5) and stayed at elevated levels as long as post-partum day 30 (PP30) (Fig. 1 C and D). We observed two pro-liferative bursts of Tregs in early (E2.5) and late (E18.5) pregnancyby Ki67 staining, coinciding with the first contact to paternal sperm

antigen and the systemic challenge with fetal antigen following pla-cental perfusion, respectively. In line with this observation, the firstexpansion was limited to the local paraaortic lymph nodes, whereasthe second was also detectable in regional inguinal lymph nodes andsystemically in the spleen (Fig. 1E). Additionally, late pregnancyTregs consistently showed increased expression of the immunosup-pressive molecule CTLA-4 in all assessed organs, rendering thempotentially more effective in controlling effector responses (Fig. 1E).CD4+Foxp3− conventional T cells (Tcons) showed only one pro-liferative burst in early pregnancy (E2.5) that was accompanied byincreased CTLA-4 expression, whereas their proliferative activityappeared to be tightly controlled at later time points (Fig. S1).Together, the Treg response commenced in early gestation

with a local proliferative burst and generalized toward systemiccompartments in late gestation. Late gestational Tregs showedincreased expression of CTLA-4, potentially supporting theirsuppressive function.

T-Cell Intrinsic Sensing of Progesterone Mediates Treg Enrichment inVitro. Progesterone is an essential steroid hormone for successfulpregnancy outcome that peaks at late gestation (18–21, 38, 39),

E2.5 E10.5 E18.5

0

20

40

60

Ki67

+ am

ong

Treg

(%)

60

70

80

90

100

CTL

A-4+

am

ong

Treg

(%)

Ki67

+ am

ong

Treg

(%)

CTL

A-4+

am

ong

Treg

(%)

Ki67

+ am

ong

Treg

(%)

CTL

A-4+

am

ong

Treg

(%)

0

20

40

60

60

70

80

90

1000

20

40

60

60

70

80

90

100

Para-aortic LN

Non-pr

egE2.5

E10.5E18

.5PP5

PP10PP30

Non-pr

egE2.5

E10.5E18

.5PP5

PP10PP30

Non-pr

egE2.5

E10.5E18

.5PP5

PP10PP30

Non-pr

egE2.5

E10.5E18

.5PP5

PP10PP30

Non-pr

egE2.5

E10.5E18

.5PP5

PP10PP30

Non-pr

egE2.5

E10.5E18

.5PP5

PP10PP30

0

5

10

15

20

25

Treg

amon

gC

D4+

cel

ls (%

)

Treg

amon

gC

D4+

cel

ls (%

)

Treg

amon

gC

D4+

cel

ls (%

)

Inguinal LN

0

5

10

15

20

Spleen

Para-aortic LN Inguinal LN Spleen

0

5

10

15

20

Para-aortic LN

BALB/c ♂

C57BL/6 ♀

E2.5Non-preg

Early Mid Late

Delivery

Pregnancy

Post partum

Progesteroneserum levels

Mating

E10.5 E18.5 PP5 PP10 PP30

×

Spleen

Para-aortic LN

Inguinal LN

Pregnant uterus

Aorta

CD4

Foxp

3

Non-preg PP5 PP10 PP30

11.9 11.8 13.6 18.9 15.116.218.3

C

A B

E

D

****

** *

**

** *

**

**

** ** *** ** ** **

*

** ** **

Fig. 1. Treg dynamics during pregnancy. Flowcytometry analysis of paraaortic LNs, inguinal LNs, andspleen at indicated pregnancy (E2.5–E18.5) and post-partum (PP5–PP30) time points. Gray shaded areasrepresent pregnancy. (A) Time course of allogeneicmating, sample collection, and progesterone serumlevels. (B) Overview of harvested lymphoid tissue inrelation to pregnant uterus. (C) Representative dotplots of Treg frequency in paraaortic LNs. (D) Quanti-fication of Treg frequency in indicated tissues. Eachdot represents the result from one individual mouse(n = 7, 8, 8, 8, 6, 6, and 4 per time point). (E) Pheno-typic characterization of Tregs. Data are pooled frommultiple experimental days. Statistical analyses in Dand E were performed by one-way ANOVA withBonferroni’s post hoc test in comparison with non-pregnant control mice. *P < 0.05; **P < 0.01.

E182 | www.pnas.org/cgi/doi/10.1073/pnas.1617115114 Engler et al.

shortly before we observed a substantial Treg expansion (Fig.1A). Additionally, progesterone has been shown to have bene-ficial effects in EAE (18, 22, 40) and to expand Treg frequenciesupon in vivo treatment (24). Therefore, we tested the ability ofprogesterone to influence the ratio of Tregs and Tcons in sple-nocyte cultures. Indeed, treatment with progesterone at dosesclose to late gestational serum levels (∼90 ng·mL–1) readily in-creased the frequency of Foxp3+ Tregs among live CD4+ T cellsfrom ∼5% in vehicle-treated cultures to ∼15% after 48 h (Fig. 2A and B). We could inhibit this effect by the steroid hormonereceptor antagonist mifepristone (RU486), suggesting a defined

receptor-mediated rather than a pleiotropic effect. Importantly,RU486 alone as well as the pregnancy hormone estradiol had noeffect on Treg frequencies (Fig. S2A). At the same time, weobserved increased cell death in progesterone-treated culturesparalleling the enrichment of Tregs in a dose-dependent manner(Fig. 2 A–C and Fig. S2B). Prolonged cultivation amplified theeffect (Fig. 2D), whereas 1 μg·ml–1 RU486 was under mostconditions sufficient for a complete blockade (Fig. 2B). Be-cause splenocyte cultures contain different immune cell typesthat could in principle be involved in sensing progesterone andmediating this finding, we sought to define the target cells of

0

5

10

15

20

25

Incubation time

Treg

am

ong

viab

le C

D4+

cel

ls (%

)C

ell d

eath

amon

g C

D4+

cel

ls (%

)Tr

eg a

mon

g vi

able

CD

4+ c

ells

(%)

Treg

am

ong

viab

le C

D4+

cel

ls (%

)

T reg

am

ong

v iab

le C

D4+

cel

ls (%

)

P4+RU486

EtOHP4

0 7 15 30 6012025

050

010

0020

000

20

40

60

80

100

P4 ng/ml

24 h 48 h 72 h 96 h0

5

10

15

20

EtOHP4P4+RU486

7.1 23.7 10.1

EtOH P4

P4+RU48

60

2

4

6

8

EtOH P4 P4+RU486

CD4

Foxp

3

Viable CD4+ cells Viable CD4+ cells Viable CD4+ cells

Cell death dye

SSC

-A

CD4+ cells CD4+ cells CD4+ cells

16.55.9 7.1

CD4

Foxp

3

EtOH P4

DMSO

Z-DEVD-FMK

CD4

Anne

xin

VaC

asp3

EtOH P4 P4+RU486

EtOH P4

P4+RU48

60

2

4

6

8

10

Anne

xinV

+ am

ong

viab

le C

D4+

cel

ls (%

)

EtOH P4

P4+RU48

60

2

4

6

aCas

p3+

amon

gvi

able

CD

4+ c

ells

(%)

7.9 9.5 7.6

2.9 5.3 3.7

EtOH P4

0

5

10

15

DMSOZ-DEVD-FMK

4.8 12.1

4.5 7.6

A B

C D E

G

F

****

***

***

******

*

****

**

**

** **** **

*

0 50 1000

5

10

15

20

Cell death among CD4+ cells (%)

Treg

am

ong

viab

le C

D4+

cel

ls (%

)

R2 = 0.8693P < 0.0001

Fig. 2. T-cell–intrinsic sensing of progesterone medi-ates Treg enrichment in vitro. (A) Splenocytes werecultured for 48 h in the presence of 300 ng·mL–1 pro-gesterone (P4), 1 μg·mL–1 mifepristone (RU486), vehiclecontrol ethanol (EtOH), or indicated combinations.Cultures were analyzed for Treg frequency and celldeath by flow cytometry. (B) Dose titration of spleno-cytes cultured and analyzed as described in A. (C) Cor-relation of Treg frequency and cell death among CD4+

cells from dose titration experiment in B. (D) Timecourse of splenocytes cultured and analyzed as de-scribed in A. (E) Purified CD4+ cells were cultured (0.2 ×106 cells per well) and analyzed as described in A.(F) Splenocytes were cultured as in A but analyzed after6 h for apoptosis markers Annexin V and aCasp3.(G) Splenocytes were cultured in the presence of cas-pase 3 inhibitor Z-DEVD-FMK or vehicle control (DMSO)and analyzed as in A. Dot plots in A are representativeof at least three independently analyzed animals. Datain B are pooled from five independent experimentswith one mouse per experiment (total n = 5). Data inC show one representative animal out of five (all ani-mals are shown Fig. S2B). Data in D show results of oneexperiment (n = 5). Data in E show one representativeexperiment out of two (each n = 4). Data in F arepooled from two independent experiments (totaln = 8). Data in G are pooled from three independentexperiments (total n = 10). Statistical analysis was per-formed by linear regression in A, two-way ANOVA inB, D, and G, and one-way ANOVA in E and F, all withBonferroni’s post hoc test. *P < 0.05; **P < 0.01.

Engler et al. PNAS | Published online January 3, 2017 | E183

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

progesterone. To get a first impression, we repeated the assaywith purified CD4+ T cells. Because the progesterone-inducedenrichment of Treg cells was preserved in these experiments (P =0.0003, Fig. 2E), we concluded a direct action of progesterone onCD4+ T cells.Because cell death was increased in progesterone-treated

cultures, we further analyzed Annexin V and activated cas-pase 3 (aCasp3) to test whether CD4+ T cells were specifi-cally driven into apoptosis. Indeed, both markers showed anincrease in apoptosis after progesterone treatment, which wasabolished in the presence of RU486 (Fig. 2F). More impor-tantly, treatment with the caspase 3 inhibitor Z-DEVD-FMKreduced Treg frequencies in progesterone-treated cultures(Fig. 2G), indicating that apoptosis was directly driving Tregenrichment.Thus, T-cell–intrinsic sensing of progesterone via a RU486-

blockable receptor resulted in a shift of the immunological balancein favor of Tregs. This effect was driven by the induction of apo-ptosis and consecutive cell death in the CD4+ T-cell compartment.

Progesterone Acts via Promiscuous Binding to the GlucocorticoidReceptor. Our data suggested the existence of a RU486-blockablereceptor in T cells that is engaged by progesterone and inducesan enrichment of Tregs in vitro. To explore potential moleculartargets of progesterone, we next performed gene expression analysesof different steroid receptors in isolated Tregs and Tcons throughoutpregnancy, including the progesterone receptor (PR encoded byNr3c3/Pgr), the glucocorticoid receptor (GR encoded by Nr3c1) andthe estrogen receptor α (ER encoded by Nr3a1/Esr1). Notably, PgrmRNA was practically absent in all conditions, whereas Nr3c1and Esr1 mRNA could be reliably detected (Fig. 3A). Becausesteroid hormones are known to possess promiscuous binding ac-tivity to other steroid receptors (41), we reasoned that progesteronemight signal via the GR as an alternative receptor. To test thishypothesis, we challenged splenocytes with either norgestrel (NOR)or dexamethasone (DEX), possessing high affinity to the PR orGR, respectively. Although applied at lower concentrations, DEXtreatment showed a much stronger enrichment of Tregs than NORtreatment (P = 0.0009, Fig. 3A), supporting our hypothesis.

0

20

40

60

80

%am

ong

viab

le s

plen

ocyt

es

T cells

B cells

Neutro

phils

Macrop

hage

s

NK cells

DCs0

20

40

60

%am

ong

viab

le s

plen

ocyt

es GRfl/flGRfl/fl;Lck-Cre

GRfl/flGRfl/fl;Lck-Cre

Treg

amon

g vi

able

CD

4+ c

ells

(%)

Cel

l dea

th a

mon

gC

D4+

cel

ls (%

)Tr

eg a

mon

g vi

able

CD

4+ c

ells

(%)

0

0.5

1.0

1.5

2.0Nr3c1

Non-preg E18.5 PP5 Non-preg E18.5 PP5 Non-preg E18.5 PP50

0.02

0.04

0.06

0.08

0.10

Rel

ativ

em

RN

Ale

vel

Rel

ativ

em

RN

Ale

vel

Rel

ativ

em

RN

Ale

vel

PgrTconTreg

TconTreg

TconTreg

2/5 5/5 3/6n.d. n.d. n.d. 0

0.05

0.10

0.15

0.20Esr1

T cells

CD4+

T cells

CD8+

T cells

Treg

EtOHNOR

NOR+RU48

6DEX

DEX+RU48

60

5

10

7.4 12.8 17.6

7.9 7.4 7.4

CD4

Foxp

3

GRfl/fl

EtOH P4 DEX

GRfl/fl;Lck-Cre

EtOH P4

P4+RU48

6DEX

DEX+RU48

6

EtOH P4

P4+RU48

6DEX

DEX+RU48

6

0

10

20

30

40

0

5

10

15

20

25

GRfl/flGRfl/fl;Lck-Cre

GRfl/flGRfl/fl;Lck-Cre

Cell death dye

SSC

-A

GRfl/fl

EtOH P4 DEX

GRfl/fl;Lck-Cre

5.7 23.0 31.0

5.2 8.2 8.7

A

B C

D

E

****

**

**

**

**

Fig. 3. Progesterone acts via promiscuous binding tothe glucocorticoid receptor. (A) Relative mRNA levelsof progesterone receptor (Pgr), glucocorticoid re-ceptor (Nr3c1), and estrogen receptor α (Esr1) in Tconsand Tregs from nonpregnant (n = 5), pregnant (E18.5;n = 5), and postpartum (PP5; n = 6) mice. mRNA wasquantified by real-time PCR and normalized to Tbp.(B) Splenocytes were cultured for 48 h in the presenceof 300 ng·mL–1 NOR, 500 pg·mL–1 (∼10−9 M) DEX,1 μg·mL–1 mifepristone (RU486), vehicle control etha-nol (EtOH), or indicated combinations. Cultures wereanalyzed for Treg frequency by flow cytometry.(C) Immune cell composition of splenocytes from T-cell–specific glucocorticoid receptor knockout mice (GRfl/fl;Lck-Cre) and controls (GRfl/fl) was analyzed by flowcytometry. (D and E) Splenocytes of GRfl/fl;Lck-Cre andGRfl/fl mice were cultured for 48 h in the presence of300 ng·mL–1 progesterone (P4), 500 pg·mL–1 (∼10−9 M)DEX, 1 μg·mL–1 mifepristone (RU486), vehicle controlethanol (EtOH), or indicated combinations. Cultureswere analyzed for Treg frequency (D) and cell death(E). Data in A are pooled from multiple experimentaldays. Data in B show results for one experiment (n = 4).Data in C–E show one representative experiment outof two (n = 4 per group). Statistical analysis was per-formed by two-way ANOVA in A and C–E and one-wayANOVA in B, all with Bonferroni’s post hoc test. *P <0.05; **P < 0.01; n.d. = not detected; x/y = number ofsamples with signal.

E184 | www.pnas.org/cgi/doi/10.1073/pnas.1617115114 Engler et al.

To definitely pinpoint a contribution of GR engagement tothe observed Treg enrichment, we made use of a T-cell–specificGR knockout mouse. In theseGRfl/fl;Lck-Cremice the T-cell–restrictedlymphocyte-specific protein tyrosine kinase (Lck) promoter drivesthe expression of a Cre recombinase, which leads to the excision ofthe essential exon 3 from the GR locus, thereby disrupting GRfunction specifically in T cells (42–44). After ruling out a prioridifferences in their immune cell composition (Fig. 3C), we cul-tured splenocytes fromGRfl/fl;Lck-Cre knockout andGRfl/fl controlanimals in the presence of either progesterone or DEX. Strikingly,Treg enrichment and cell death induction were completely abol-ished in the GRfl/fl;Lck-Cre knockout cultures, regardless of treat-ment with either progesterone or DEX (Fig. 3 D and E). Thiscomplete rescue in the GR knockout cultures also applied to theinduction of apoptosis (Fig. S3 A and B), whereas Treg enrichmentby both progesterone and DEX was still readily observed in pro-gesterone receptor knockout cultures (Fig. S3C). In line with ourprevious experiments, these data support a direct action of pro-gesterone on the GR in T cells to mediate Treg enrichment invitro. To test the effect of GR engagement on T cells in vivo, wetreated wild-type animals with a single i.p. injection of 100 μg DEXand analyzed lymph nodes and spleen 24 h later. Markedly, Tregfrequencies were strongly increased by DEX in all assessed organs(Fig. S4A), whereas this effect was blocked in the GRfl/fl;Lck-Cremice (Fig. S4B).We conclude that Treg enrichment upon steroid challenge was

mediated via the GR in T cells, whereas we found no indicationfor an involvement of the PR.

Tregs Are More Resistant to GR Challenge than Tcons. Becauseprogesterone treatment was paralleled by increased cell death inCD4+ T cells (Fig. 2 A and B), the increased Treg-to-Tcon ratioobserved in vitro could be explained by an increased steroid sen-

sitivity of Tcons in comparison with Tregs. For example, it is con-ceivable that Tcons are more sensitive to glucocorticoid-inducedcell death, whereas Tregs are in comparison steroid resistant andhence accumulate. This notion was supported by our observationthat cell death dye-positive cells were exclusively present in theTcon but not in the Treg compartment (Fig. 4A). To further sub-stantiate this hypothesis, we isolated Tregs and Tcons and culturedthem separately in the presence of increasing doses of DEX as apotent GR stimulus. Whereas we observed more robust proliferativeresponses of Tcons at low concentrations of DEX (P = 0.0380),increasing GR stimulation resulted in a proliferative advantage forTregs (P = 0.0017, Fig. 4B). Similarly, induction of cell death wasmore pronounced in Tcons than in Tregs when cultured undernonproliferating conditions (P < 0.0001, Fig. 4C). These findingssupport the idea that differential steroid sensitivity in the CD4+

T-cell compartment accounts for an enrichment of the more re-sistant Tregs in situations of increased GR activity.To check whether steroid levels during pregnancy are sufficient

to induce transcriptional activity of the GR in CD4+ T cells in vivo,we isolated Tregs and Tcons from nonpregnant and E18.5 miceand analyzed gene expression of the glucocorticoid-induced leu-cine zipper (Gilz), encoded by Tsc22d3. We selected Gilz becauseits transcription is directly induced by binding of the GR to theGilz promoter (45, 46), it mediates the antiproliferative activity ofglucocorticoids (47), and can be used as a surrogate marker forGR activity. Gilz expression was significantly induced by latepregnancy in Tcons (P = 0.0207) but not in Tregs (P = 0.2496,Fig. 4D). Additionally, GR signaling was more active in Tconsirrespective of pregnancy, further supporting the notion that thosecells might be a priori more susceptible to steroid challenge.Taken together, Tregs showed a higher resistance to glucocor-

ticoid challenges under both proliferative and nonproliferativeconditions. This finding is consistent with a positive selection of

−12 −11 −10 −9 −8 −7 −60

0.5

1.0

log M DEX

Rel

ativ

epr

olife

ratio

n TconTreg

10−8 M DEX10−9 M DEXEtOH

10−8 M DEX+RU486

0

1

2

3

Cel

ldea

th fo

ld c

hang

e

4 h incubation time

0

1

2

36 h incubation time

Cel

ldea

th fo

ld c

hang

e

Curve bottom valueCurve top value

Tcon Treg0

0.05

0.10

0.15

0.20

0.25

Rel

ativ

epr

olife

ratio

n

Tcon Treg Tcon Treg

Tcon Treg0.7

0.8

0.9

1.0

1.1

1.2

Rel

ativ

epr

oli fe

ratio

n

Tsc22d3

EtOH P4 P4+RU486

Tcon

Treg

Foxp3

Rel

ativ

e m

RN

Ale

vel

1.47 2.77 1.77

32.47.2 9.7

Cel

l dea

th d

ye

0 7 15 30 6012025

050

010

0020

000

20

40

60

80

100

P4 ng/ml

Cel

ldea

th (%

)

TconTreg

Tcon Treg0

1

2

3

4

5 non-pregpreg

B

A

C D

**

*****

***

****

******

*

Fig. 4. Tregs are more resistant to GR challenge thanTcons. (A) Splenocytes were cultured for 48 h in thepresence of 300 ng·mL–1 progesterone (P4), 1 μg·mL–1

mifepristone (RU486), vehicle control ethanol (EtOH),or indicated combinations. Cultures were analyzed forcell death dye positivity inside Tcon and Treg pop-ulations. (B) Proliferation response curves of Tregs andTcons. Cells were isolated and cultured separately inthe presence of irradiated feeder cells, 1 μg·mL–1 anti-CD3 antibody, 50 U·mL–1 recombinant murine IL-2,and indicated DEX concentrations. Proliferation wasassessed by [3H]-thymidine incorporation and nor-malized to vehicle control (EtOH). Curve fit, curve top,and curve bottom values were computed. (C) Tregsand Tcons were isolated and cultured separately in thepresence of DEX, 1 μg·mL–1 mifepristone (RU486), ve-hicle control ethanol (EtOH), or indicated combina-tions. Dead cells were identified by propidium iodidepositivity and cell death was calculated as fold changerelative to vehicle control (EtOH). (D) Relative mRNAlevels of Gilz (encoded by Tsc22d3) in Tcons and Tregsfrom nonpregnant (n = 5) and pregnant (E18.5; n = 5)mice. mRNA was quantified by real-time PCR andnormalized to Tbp. Data in A are reanalyzed from Fig.2B (total n = 5). Data in B are pooled from three in-dependent experiments (total n = 4). Data in C arepooled from three independent experiments (totaln = 3). Data in D are pooled from multiple experi-mental days (n = 5 per group). Statistical analyses wereperformed by two-way ANOVA with Bonferroni’s posthoc test in A, C, and D and Student’s t test in B. *P <0.05; **P < 0.01.

Engler et al. PNAS | Published online January 3, 2017 | E185

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

Tregs as the underlying mechanism for increased Treg frequenciesafter GR stimulation. The GR pathway appeared to be engagedduring pregnancy in vivo, because the GR target gene Gilz wasinduced in Tcons at E18.5.

Late Pregnancy Temporarily Protects from CNS Autoimmunity. Next,we aimed to unravel whether GR activity in T cells is responsiblefor the beneficial effect of pregnancy on CNS autoimmunity. Tothis end, we first established a model of pregnancy protection fromEAE in C57BL/6 mice. We performed allogeneic matings withBALB/c males and immunized at gestational day E7.5 ± 1, en-suring an estimated onset of disease at the beginning of the thirdtrimester of pregnancy—the time point with the strongest pro-tective effects in MS and EAE (9, 13). We expected the delivery ofthe pups around day 12 after immunization and included non-pregnant females as controls (Fig. 5A). Whereas the disease in-cidence of both groups was indistinguishable (P = 0.5681, Fig. 5B),pregnant animals were protected until the day of delivery. Imme-diately after birth of the pups, the dams succumbed to an exag-gerated disease course associated with increased motor disabilityand higher mortality, reminiscent of the overshooting disease ac-tivity observed in MS patients postpartum (Fig. 5 C–E) (9, 11).Because motor function defects assessed during EAE can be

attributed to inflammatory injury of responsible neurons and theirprojections (48, 49), we next assessed the infiltration of immunecells into the CNS during the protective period (at day 10) by flowcytometry. At this time point, pregnant animals were clinicallyasymptomatic (Fig. 5F) and showed half the absolute number ofimmune cells in the CNS (P = 0.0089, Fig. 5G), suggesting a pe-ripheral control of encephalitogenic cells, whereas particularlyT cells and dendritic cells (DCs) were reduced in pregnant EAEbrains (Fig. 5H).

The GR in T Cells Mediates EAE Protection During Pregnancy but IsDispensable for Reproductive Success. To assess the biological roleof GR in T cells in mediating the protective effect of pregnancy in

CNS autoimmunity, we performed allogeneic matings of GRfl/fl;Lck-Cre knockout andGRfl/fl control animals and induced EAE ongestational day E7.5 ± 1. We observed no differences in the dis-ease incidence (Fig. 6A). However, as expected from our initialexperiments, pregnancy resulted in a protection of GRfl/fl controlmice. By contrast, the EAE protection was abolished in pregnantGRfl/fl;Lck-Cre knockout mice (P = 0.003, Fig. 6B), whereas wefound no evidence for an a priori exacerbated EAE in GRfl/fl;Lck-Cre knockout animals when comparing the nonpregnant groups(Fig. 6 B–D). Thus, protection from EAE in pregnant micedepended on the presence of the GR in T cells. Additionally, Tregfrequencies in pregnant GRfl/fl;Lck-Cre knockout animals withEAE (E18.5) were reduced in comparison with pregnant EAEcontrols in paraaortic lymph nodes (P = 0.0144, Fig. 6E), sug-gesting an impaired Treg expansion. Although pregnancy stillshowed a tendency to increase Tregs in GRfl/fl;Lck-Cre mice, thisdid not reach statistical significance (P = 0.1023, Fig. 6E).To investigate the possibility that a failure in fetomaternal tol-

erance with consecutive fetal loss could have caused this break-down of EAE protection as a secondary effect, we killed GRfl/fl;Lck-Cre mice and controls at E13.5–14.5 and analyzed re-productive outcome parameters, including fetal loss and pla-cental morphology. However, all assessed parameters, includingTreg frequencies, were unchanged in GRfl/fl;Lck-Cre mice (Figs.S5 and S6). Thus, fetomaternal tolerance appeared to be morestable than pregnancy-induced autoantigen tolerance.

DiscussionIn the present study, we have in detail characterized shifts in theT-cell compartment during pregnancy with a specific focus onthe immunological balance of Tcons and Tregs and the endo-crine mechanisms orchestrating them. Our findings indicate thatpregnancy-related steroid hormones are directly sensed byT cells and mediate a relative enrichment of Tregs via engage-ment of the GR. Intriguingly, our conditional GR knockoutmodel demonstrated that steroid hormone sensing by T cells is

BALB/c ♂

C57BL/6 ♀

EAEinduction

Non-pregnantC57BL/6 ♀

PregnantE7.5 ± 1

Day 0 Day 7

ExpectedEAE onset

Day 12

Expecteddelivery

Thirdtrimester

Delivery

5 10 15 20 250

1

2

3

4

Days after immunization

Mea

n cl

inic

al s

core

PregNon-preg

Non-preg Preg0

5

10

15

20

Day

of o

n set

PregNon-preg

Dis

ease

inci

denc

e

Non-preg Preg0

0.2

0.4

0.6

0.8

1.0

0 5 10 15 20 250

20

40

60

80

100

Days after immunization

Perc

ent s

urvi

val

Microg

lia

Neutro

phils

NK cells

DCs

Macrop

hage

s

B cells

T cells

0

1

2

3

T cells

CD4+

T cells

CD8+

T cells

Treg

0

1

2

Non-pr

egPreg

Non-pr

egPreg

0

2

4

6

8

10

Cum

ulat

ive

clin

ical

sco

re

0

0.5

1.0

1.5

Imm

une

cells

in C

NS

(× 1

06)

Imm

une

cells

in C

NS

(× 1

05)

Imm

une

cells

in C

NS

(× 1

05)

Non-pregPreg

Non-pregPreg

×

A B

C D E

F G H

**

*

**

****

*** **

Fig. 5. Late pregnancy temporarily protects fromCNS autoimmunity. (A) Experimental setup of preg-nancy EAE. Disease incidence (B), clinical course (C),day of onset (D), and survival (E) are shown fornonpregnant (n = 17) and pregnant (n = 12) animals.Gray shaded areas represent pregnancy. (F–H) Non-pregnant (n = 4) and pregnant (n = 8) animals werekilled on day 10 after EAE induction. Cumulativeclinical score (F) and absolute numbers of immunecells (G) and immune cell subpopulations in the CNS(H) were assessed by flow cytometry. Data in B–E arepooled from four independent experiments. Data inF–H show one representative experiment out oftwo. Statistical analyses were performed by Fisher’sexact test in B, two-way ANOVA with Bonferroni’spost hoc test in C and H, Student’s t test in D, F, andG, and χ2 test in E. *P < 0.05; **P < 0.01.

E186 | www.pnas.org/cgi/doi/10.1073/pnas.1617115114 Engler et al.

crucial for conferring protection from maternal autoimmunity inlate gestation. However, establishment of fetomaternal toleranceitself, reflected by reproductive success, did not depend on thepresence of GR in T cells.Although Treg-driven tolerance has been shown to be fetal

antigen specific in the first place (22), it has also been noted thatTregs are able to suppress T-cell responses to unrelated antigensin the form of a bystander suppression (50). This can have detri-mental consequences as Treg expansion leaves pregnant micemore prone to infections with certain pathogens including Listeriaand Salmonella species, whereas Treg depletion has been shownto restore host defense (23). Thus, enhanced Treg-driven controlof effector T-cell responses in pregnancy represents a promisingtarget when studying tolerogenic off-target effects, as it might bethe case in pregnancy-induced reduction of autoimmunity.Herein, we characterized Tregs and Tcons in C57BL/6 females

mated to BALB/c males and observed a marked late gestationalTreg expansion in the local paraaortic LNs that directly drain theuterus and constitute the place of fetal-specific T-cell priming(2, 51). Similar findings have been reported by other groups pio-neering the investigation of Tregs in the context of reproductivebiology (18, 23, 52). Notably, increased Treg frequencies persistedeven after delivery. This finding is in concordance with a studyusing a mating system with known paternal antigen to show thatfetal-specific Tregs can persist as regulatory memory cells thatreadily reassemble in a consecutive pregnancy (22). Beyond that,we identified proliferative bursts in Tregs that appear to reflect alocal priming event driven by male seminal fluid (51) in earlygestation and a systemic challenge to fetus-derived antigens shedinto the mother’s circulation (52, 53) in late gestation.However, the most dramatic changes in Treg frequency and

phenotype at E18.5 were directly preceded by the peak pro-gesterone levels present around E16.5 (38). A number of studieshave implied that progesterone can support Tregs and dampeneffector responses, without providing a clear molecular mecha-nism. These include observations that s.c. progesterone-releasingimplants at the onset of disease ameliorate the course of EAE inC57BL/6 mice, however, without altering the systemic frequencyof Tregs (40). In contrast, another study reported a substantialincrease of uterine and systemic Treg frequencies when treatingovariectomized mice with s.c. injection of progesterone (24).

Additionally, progesterone was shown to drive naïve human cordblood cells but not adult peripheral T cells into suppressive Tregswhile impeding their differentiation into TH17 cells (25).Thus, we addressed the questions of whether progesterone is

capable of shifting the balance in the T-cell compartment in favor ofTregs and whether we could set up a stable assay to decipher theunderlying molecular mechanism. Indeed, we observed a robustincrease of Treg frequencies after treating splenocyte cultures for48 h with 300 ng·mL–1 progesterone, which corresponds to ap-proximately three times the serum level present at E16.5 (38). Weconsider this level close to the physiological range, as we wouldanticipate similar local levels close to the fetomaternal interface.To our surprise however, we found no evidence for PR expres-

sion in CD4+ T cells on the mRNA level, representing the primaryprogesterone target. This finding is in agreement with data from theImmunological Genome Project (54) providing Pgr expression levelsthat do not exceed the detection background. By a series of ex-periments, we could finally attribute the effect to a cross engage-ment of the GR in T cells. The strongest piece of evidence in thisregard is the complete abrogation of Treg enrichment in culturesfrom T-cell–restricted GR knockout mice (GRfl/fl;Lck-Cre). Thisevidence suggests that T cells sense progesterone via their in-tracellular GR, leading to an increase of Treg frequencies. Beingengaged in times of increased steroid serum levels, this finding rep-resents a so-far unrecognized tolerogenic pathway that might shapethe immunological balance during pregnancy. Clearly, this mecha-nism is not restricted to progesterone, but rather includes the actionof corticosterone and other steroid hormones with agonistic potentialon the GR that equally increase during the course of gestation (55).Mechanistically, we found that Tregs are more resistant to

steroid challenges in comparison with Tcons, thus possessing asurvival advantage in times of increased steroid levels. Hence, theobserved enrichment of Tregs is likely the result of a selection ofresistant Tregs, whereas sensitive Tcons succumbed to glucocor-ticoid-induced cell death. This interpretation is also in agreementwith increasing rates of apoptosis and cell death in CD4+ cellsparalleling the Treg enrichment in dose and time-course experi-ments. Importantly, targeted inhibition of apoptosis reduced theTreg enrichment, further supporting this notion. Whereas apoptosisinduction in leukocytes by DEX is well established (56), there haveonly been incidental reports for progesterone (57). Furthermore,

Treg

amon

gC

D4+

cel

ls (%

)

GRfl/fl

GRfl/fl;Lck-CreGRfl/fl preg

GRfl/fl;Lck-Cre preg

GRfl/fl

GRfl/fl;Lck-CreGRfl/fl preg

GRfl/fl;Lck-Cre preg

GRfl/fl

GRfl/fl;Lck-CreGRfl/fl preg

GRfl/fl;Lck-Cre preg

0

10

20

30Para-aortic LN Inguinal LN Spleen

0

0.2

0.4

0.6

0.8

1.0

Dis

ease

inci

denc

e

5 10 150

1

2

3

4

Days after immunization

Mea

ncl

inic

alsc

ore

0 5 10 150

50

100

Days after immunization

Perc

ent s

urvi

val

0

5

10

15

20

25

Day

of o

nset

GRfl/fl

GRfl/fl ;Lck-Cre

GRfl/fl pre

g

GRfl/fl ;Lck-Cre

preg

0

10

20

30

0

10

20

30

GRfl/fl

GRfl/fl;Lck-CreGRfl/fl preg

GRfl/fl;Lck-Cre preg

Para-aortic LNNon-preg Preg

CD4

Foxp

3

GRfl/fl

GRfl/fl;Lck-Cre

9.8 18.2

10.9 14.3

A

E

B C D

*

**

Fig. 6. The GR in T cells mediates EAE protectionduring pregnancy. (A–D) EAE was induced in non-pregnant and pregnant GRfl/fl (n = 20 and n = 18, re-spectively) and nonpregnant and pregnant GRfl/fl;Lck-Cre mice (n = 21 and n = 14, respectively). Diseaseincidence (A), clinical course (B), day of onset (C), andsurvival (D) are shown. Gray shaded areas representpregnancy. (E) Treg frequency in nonpregnant andpregnant GRfl/fl (n = 10 and n = 14, respectively) andnonpregnant and pregnant GRfl/fl;Lck-Cre mice (n = 5and n = 16, respectively) treated as in A–D were killedat E18.5 and analyzed by flow cytometry. Data arepooled from four independent experiments in A–Dand three independent experiments in E. Statisticalanalyses were performed by Fisher’s exact test in A;two-way ANOVA in B; one-way ANOVA in C and E, allwith Bonferroni’s post hoc test; and χ2 test in D. *P <0.05; **P < 0.01.

Engler et al. PNAS | Published online January 3, 2017 | E187

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

we observed increased Gilz expression in Tcons isolated from preg-nant animals, indicating that GR activity is induced during pregnancyin these cells, but to a lesser extent in the more resistant Tregs.To investigate the relevance of this mechanism for the ame-

lioration of autoimmunity during in vivo pregnancy, we estab-lished a mouse model in which pregnant animals are protectedfrom EAE until delivery. Strikingly, whereas the protectionconferred by pregnancy is among the strongest beneficial regu-lators of autoimmunity, this effect was completely abrogatedin GRfl/fl;Lck-Cre knockout animals, supporting a central roleof GR signaling in ensuring proper maintenance of pregnancy-related CNS-autoantigen tolerance. Accordingly, the Treg expansionin paraaortic LNs present in pregnant controls was diminished inpregnant GRfl/fl;Lck-Cre knockout animals during EAE. Our resultsextend previous evidence supporting the crucial role of intact GRsignaling in T cells for the control of neuroinflammation (58). It hasbeen also reported that deletion of the GR in T cells abolishes thetherapeutic effect of dexamethasone in mice (59, 60), whereasoverexpression of the GR in thymocytes ameliorates EAE in a ratmodel (61). Here, we show that pregnancy endogenously employsthis pathway to exert its protective effects on maternal autoimmunity.Additionally, we provide evidence that the GR in T cells is of

particular importance for mediating amelioration of maternal au-toimmune responses but not for ensuring fetomaternal tolerance asdeletion of the GR in T cells completely abrogated the protectiveeffect of pregnancy on EAE but had no impact on any assessedmeasures of reproduction. Notably, also Treg expansion found tobe disturbed in GRfl/fl;Lck-Cre knockout animals in EAE preg-nancy was unaltered in mid term pregnancy without EAE. Thesefindings likely reflect the high stability of reproduction and Tregexpansion, which are ensured by redundant mechanisms in theabsence of inflammatory challenges (1, 5). In line with this re-dundancy, it usually takes more than one hit to cause severe re-productive failure, as evidenced by a number of mutant mice withlargely unaltered reproductive success (1). This result is likely dueto high evolutionary pressure on successful species propagation viareproduction (62). In contrast, controlling chronic inflammation inautoimmune diseases seems to be a more fragile side effect thatrequires all reproductive tolerance mechanisms to fully operate.Some limitations of this study have to be considered. Whereas

our data clearly demonstrate that GR signaling in T cells is in-dispensable for pregnancy protection from EAE, this does not ruleout additional protective mechanisms of pregnancy exerted byother factors such as estrogens, which have shown promise inclinical trials in MS (63–65). Furthermore, because we show thatsuccessful reproduction does not critically rely on GR signaling inT cells, it needs to be further elucidated on what basis differentialeffects on reproductive and autoimmune tolerance are mediated.Summarizing our findings, we provide evidence that differential

steroid sensitivity of Tregs and Tcons represents a so-far un-recognized tolerance mechanism that might be engaged in timesof high steroid levels, as present during gestation (Fig. 7). Weshow that this pathway works via promiscuous binding of pro-gesterone to the GR in T cells and is able to enrich Tregs in vitro.Furthermore, we found evidence that GR signaling is in fact op-erative in Tcons during pregnancy in vivo, whereas T-cell–specificGR deletion resulted in a loss of pregnancy-induced protectionfrom EAE and reduced Treg frequencies in uterus-draining LNs.However, further work is needed to disentangle the molecularbasis of the differential sensitivity of Tregs and Tcons. Betterunderstanding the T-cell population-specific prerequisites thatcommunicate this differential sensitivity holds the promise to yieldmore specific therapeutic means to steer the immunological bal-ance in transplantation, cancer, and autoimmunity.

Materials and MethodsMice. C57BL/6 wild-type mice (The Jackson Laboratory) and previously de-scribed GRfl/fl;Lck-Cre (Nr3c1tm2GSc;Tg(Lck-cre)1Cwi) mice (42–44) and Pgr−/− mice

(66) were kept under specific pathogen-free conditions in the Central Ani-mal Facility at the University Medical Center Hamburg-Eppendorf. Age- andsex-matched adult animals (10–20 wk) were used in all experiments.

Dexamethasone Treatment. Mice were treated with one i.p. injection of100 μg (∼5 mg per kilogram of body weight) dexamethasone (FortecortinInject, MerckSerono) in PBS or vehicle control (PBS) and killed 24 h later.

Allogeneic Mating. Age-matched female mice were primed with housingmaterial from fertile BALB/c males for 1 wk and then mated for three con-secutive nights. Successfullymated femaleswhere identified by thepresence ofa vaginal plug, separated, andweighted daily to confirmpregnancy. The day ofplug was considered gestational day 0.5 (E0.5). In experiments includingpostpartum time points, pups were separated and killed at the day of delivery.

EAE Induction. Mice were immunized s.c. with 200 μg MOG35–55 peptide(Schafer-N) in complete Freund’s adjuvant (Difco) containing 4 mg·mL–1

Mycobacterium tuberculosis (Difco). In addition, 200 ng pertussis toxin (Calbio-chem) was injected i.v. on the day of immunization and 48 h later. Animals werescored daily for clinical signs by the following system: 0, no clinical deficits; 1, tailweakness; 2, hind limb paresis; 3, partial hind limb paralysis; 3.5, full hind limbparalysis; 4, full hind limb paralysis and forelimb paresis; and 5, premorbid ordead. Animals reaching a clinical score ≥4 had to be killed according to theregulations of the Animal Welfare Act. The last observed score of euthanized ordead animals was carried forward for statistical analysis. The cumulative clinicalscore represents the sum of the daily scores given to an animal over time. In-vestigators were blinded for genotype during the experiment.

Immune Cell Isolation from Lymphoid Organs and CNS. Mice were killed byinhalation of CO2 and lymph nodes and spleens were harvested with sterileinstruments into ice-cold PBS. Single-cell suspensions were prepared by ho-mogenization through a 40-μm cell strainer, cells were pelleted by centri-fugation (300 × g, 10 min, 4 °C), and splenic erythrocytes were lysed inred blood cell lysis buffer (0.15 M NH4Cl, 10 mM KHCO3, 0.1 mM Na2EDTA,pH = 7.4) for 5 min at 4 °C. Cells were washed with PBS and used in follow-upapplications. For isolation of CNS-infiltrating leukocytes, mice were in-tracardially perfused with ice-cold PBS immediately after killing to removeblood from intracranial vessels. Brain and spinal cord were prepared withsterile instruments, minced with a scalpel, and incubated with agitation inRPMI medium 1640 (PAA) containing 1 mg·mL–1 collagenase A (Roche) and0.1 mg·mL–1 DNaseI (Roche) for 60 min at 37 °C. Tissue was triturated

TregTcon

GR engagement

Selection ofresistant Treg

Relative Tregenrichement

GR-deficient T cellsWildtype T cells

TregTcon

No GR engagement

No selection

No Treg enrichment

EAE pregnancy in mice with GR-deficient T cells

Impaired Treg expansionSuccessful reproduction Loss of protectionfrom CNS autoimmunity

Fig. 7. Differential steroid sensitivity as a mechanism of tolerance in-duction. Upon GR engagement steroid-resistant Tregs are positively selectedand hence accumulate. After targeted GR disruption in T cells, Tcons andTregs are equally resistant to GR engagement, thus the enrichment of Treg isabolished. In pregnant EAE animals harboring GR-deficient T cells, re-production is still functional. However, Tregs increase and pregnancy-asso-ciated protection from autoimmune disease activity is impaired.

E188 | www.pnas.org/cgi/doi/10.1073/pnas.1617115114 Engler et al.

through a 40-μm cell strainer and washed with PBS (300 × g, 10 min, 4 °C).Homogenized tissue was resuspended in 30% isotonic Percoll (GE Health-care) and carefully underlaid with 78% isotonic Percoll. After gradientcentrifugation (1,500 × g, 30 min, 4 °C) CNS-infiltrating immune cells wererecovered from the gradient interphase and washed twice in ice-cold PBSbefore staining for flow cytometry.

Flow Cytometry Analysis. Single-cell suspensions were stained in the presenceof TruStain Fc receptor block (BioLegend) with monoclonal antibodiesdirected against CD3e (PerCP-Cy5.5, BioLegend, clone 145–2C11), CD4 (PE-Cy7, eBioscience, clone GK1.5), CD8a (Pacific blue, BioLegend, clone 53–6.7), CD11b (FITC, BioLegend, clone M1/70), CD11c (PE-Cy7, eBioscience,clone N418), CD25 (Alexa Fluor 488 and APC, eBioscience, clone PC61.5),CD45 (APC-Cy7, BioLegend, clone 30-F11), CD45R (V500, BD Horizon, cloneRA3-6B), Ctla4 (APC, eBioscience, clone UC10-4B9), Foxp3 (PE and eFluor450, eBioscience, clone FJK-16s), Ly6G (V450, BD Biosciences, clone 1A8),NK1.1 (PE, eBioscience, clone PK136), aCasp3 (FITC, BD Biosciences, cloneC92-605), and Ki67 (PE, BD Biosciences, clone B56) and acquired on a LSR IIFACS analyzer (BD). In indicated experiments, dead cells were labeled be-fore surface staining with Fixable Dead Cell Stain Kit (Life Technologies).For intracellular staining of Foxp3, Ctla4, and Ki67, the Foxp3 StainingBuffer Set (eBioscience) was used according to the manufacturer’s in-structions. For detection of apoptosis, the Active Caspase-3 Apoptosis Kit(BD Biosciences) and FITC Annexin V Apoptosis Detection Kit (BioLegend)were used according to the manufacturer’s instructions. The absolute cellcounts of CD45+ CNS-infiltrating leukocytes were determined by usingTrueCount tubes (BD Biosciences). Data analysis was performed withFlowJo v10 analysis software (TreeStar) for Macintosh. Tregs were identi-fied as CD4+Foxp3+ and Tcons as CD4+Foxp3– cells. Additional immune cellpopulations from splenocytes and CNS-infiltrating cells were identified aspreviously reported (49).

Isolation of T-Cell Populations. Pooled single-cell suspensions from inguinal,axillary, brachial, paraaortic lymph nodes, and spleen were subjected to mag-netic-associated cell sorting (MACS) using the CD4+ T Cell Isolation Kit (MiltenyiBiotec) or CD4+CD25+ Regulatory T Cell Isolation Kit (Miltenyi Biotec) accordingto the manufacturer’s instructions. Purity of isolated cells was routinely above80% for CD4 T cells and above 90% for Tregs and Tcons.

Steroid Assay. Isolated splenocytes were cultured in complete medium (RPMI1640, 10% FCS, 50 μM 2-mercaptoethanol, 100 U·mL–1 penicillin/streptomycin)at a cell density of 0.5 × 106 cells per well in a 96-well round bottom plate for aperiod of 48 h, unless otherwise stated. Progesterone, DEX, D(–)norgestrel,estradiol (E2), mifepristone (RU486), or vehicle control ethanol (all Sigma) wereadded at indicated concentrations. In selected experiments, cells were pre-incubated with 10 μM caspase 3 inhibitor Z-DEVD-FMK (BD Biosciences) for 30min. Four to six replicate wells were pooled per condition and subjected toantibody staining for flow cytometry analysis of viable cells.

Apoptosis Assay. Isolated splenocytes were cultured in complete medium at acell density of 0.5 × 106 cells per well in a 96-well round bottom plate for aperiod of 6 h. Progesterone, DEX, mifepristone (RU486), or vehicle controlethanol (all Sigma) were added at indicated concentrations. Four to sixreplicate wells were pooled per condition and subjected to Annexin V andaCasp3 staining for flow cytometry analysis of viable cells.

Proliferation Assay. T-cell–depleted nonproliferative feeder cells were gen-erated by adding 100 μL LowTox Rabbit Complement M (Cedarlane) and 5 μLof rat anti-mouse CD90.2 (BioLegend, clone30-H12) to 30 × 106 wild-typesplenocytes resuspended in 900 μL HBSS medium (PAA). After incubationwith mild agitation for 30 min at 37 °C, cells were washed twice with ice-coldHBSS (300 × g, 10 min, 4 °C), irradiated with 35 Gy in a Biobeam 2000 gammairradiator (Eckert & Ziegler), washed twice with ice-cold HBSS, and used asirradiated feeder cells. Tregs and Tcons were isolated by MACS as describedabove and cultured separately in complete medium at 5 × 104 cells per wellin the presence of 5 × 104 irradiated feeder cells per well in 96-well roundbottom plates. Cultures were stimulated for proliferation with 1 μg·mL–1

hamster anti-mouse CD3 antibody (BioLegend, clone 145–2C11) in thepresence of 50 U·mL–1 recombinant murine IL-2 (eBioscience) and increasingconcentrations of dexamethasone. All experimental conditions were platedin technical triplicates and adjusted to a final volume of 200 μL per well.After 48 h, cells were pulsed with 1 μCi [3H]-thymidine (Amersham) perwell for 16 h. Then, cells were harvested and spotted on filter mats with aHarvester 96 MACH III M (Tomtec). Incorporated activity per well wasassessed in counts per minute (cpm) in a beta counter (1450 Microbeta,Perkin-Elmer) and proliferation was calculated relative to the mean ofnondexamethasone-treated control wells. Curve fit and curve statisticswere computed using Prism 6 software (Graphpad) for Macintosh.

Cell Death Assay. Tregs and Tcons were isolated by MACS and culturedseparately in complete medium at 1 × 105 cells per well in 96-well roundbottom plates. Dexamethasone, mifepristone (RU486), or vehicle controlethanol were added at indicated concentrations. Dead cells were identifiedby positivity for propidium iodide (BioLegend) after 4 h and 6 h culture time.Cell death fold change was calculated relative to vehicle control.

Gene Expression Analysis. RNA was purified using RNeasy Mini Kit (Qiagen)and reverse transcribed to cDNA with RevertAid H Minus First Strand cDNASynthesis Kit (Fermentas) according to the manufacturer’s instructions. Geneexpression was analyzed by real-time PCR performed in an ABI Prism 7900HT Fast Real-Time PCR System (Applied Biosystems) using TaqMan GeneExpression Assays (Life Technologies) for the following targets: estrogenreceptor (Esr1, Mm00433149_m1), GR (Nr3c1, Mm00433832_m1), PR (Pgr,Mm00435628_m1), Gilz (Tsc22d3, Mm01306210_g1), and TATA-box bindingprotein (Tbp, Mm00446971_m1). Gene expression was calculated as 2−ΔCT

relative to Tbp as endogenous control.

Histomorphological Analyses of Placental Tissue. Paraffin-embedded placentaltissue was cut into 3- to 6-μm histological sections at the midsagittal plane usinga microtome (Leica). Tissue sections were deparaffinized, rinsed in distilledwater, and dehydrated twice in ethanol (70%). Masson-Goldner TrichromeStaining Kit (VWR International) was used according to the manufacturer’s in-structions to visualize the morphologically different areas of placental tissue.Briefly, tissue sections were stepwise stained with Weigert’s iron hematoxylin,azophloxine staining solution, phosphotungstic acid orange G, and light-greenSF solution. Finally, the tissue was dehydrated and mounted using Eukitt me-dium (O. Kindler). Image acquisition was performed using a slide scanner (MiraxMidi, Zeiss). Areas of junctional zone and labyrinth zones were quantified usingthe program MiraxViewer.

Statistics. Data were analyzed using Prism 6 software (GraphPad) for Mac-intosh and are presented as mean values ± SEM. All n values refer to thenumber of biological replicates, that is, individual mice. Differences betweentwo experimental groups were determined by unpaired, two-tailed Stu-dent’s t test. Comparison of three or more groups was performed by one-way analysis of variance (ANOVA) with Bonferroni’s post hoc test. Statisticalanalysis comparing two groups under multiple conditions or over time wasperformed by two-way ANOVA with Bonferroni’s post hoc test. In two-wayANOVA analysis of EAE disease courses, the P value for group × time in-teraction is reported. Differences in disease incidence and survival wereassessed by Fisher’s exact test and χ2 test, respectively. Significant results areindicated by asterisks: *P < 0.05; **P < 0.01.

Study Approval. All animal care and experimental procedures were carriedout according to institutional guidelines and conformed to requirements ofthe German Animal Welfare Act. Ethical approvals were obtained from theState Authority of Hamburg, Germany (approval G12/012).

ACKNOWLEDGMENTS. We thank Kristin Thiele for breeding and providingthe Pgr−/− mice. This work was supported by grants (to M.A.F. and S.M.G.)from the Forschungs- und Wissenschaftsstiftung Hamburg and DeutscheForschungsgemeinschaft (FR1720/8-1 and GO1357/8-1, KFO296 FetomaternalImmune Cross-Talk).

1. Arck PC, Hecher K (2013) Fetomaternal immune cross-talk and its consequences for

maternal and offspring’s health. Nat Med 19(5):548–556.2. Collins MK, Tay C-S, Erlebacher A (2009) Dendritic cell entrapment within the preg-

nant uterus inhibits immune surveillance of the maternal/fetal interface in mice. J Clin

Invest 119(7):2062–2073.3. Blois SM, et al. (2007) A pivotal role for galectin-1 in fetomaternal tolerance. Nat Med

13(12):1450–1457.

4. Nancy P, et al. (2012) Chemokine gene silencing in decidual stromal cells limits T cell

access to the maternal-fetal interface. Science 336(6086):1317–1321.5. Erlebacher A (2013) Mechanisms of T cell tolerance towards the allogeneic fetus. Nat

Rev Immunol 13(1):23–33.6. de Man YA, Dolhain RJEM, van de Geijn FE, Willemsen SP, Hazes JMW (2008) Disease

activity of rheumatoid arthritis during pregnancy: Results from a nationwide pro-

spective study. Arthritis Rheum 59(9):1241–1248.

Engler et al. PNAS | Published online January 3, 2017 | E189

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

7. Förger F, et al. (2008) Pregnancy induces numerical and functional changes ofCD4+CD25 high regulatory T cells in patients with rheumatoid arthritis. Ann Rheum Dis67(7):984–990.

8. Buchel E, Van SteenbergenW, Nevens F, Fevery J (2002) Improvement of autoimmunehepatitis during pregnancy followed by flare-up after delivery. Am J Gastroenterol97(12):3160–3165.

9. Vukusic S, et al.; Pregnancy in Multiple Sclerosis Group (2004) Pregnancy and multiplesclerosis (the PRIMS study): Clinical predictors of post-partum relapse. Brain 127(Pt 6):1353–1360.

10. Confavreux C, Hutchinson M, Hours MM, Cortinovis-Tourniaire P, Moreau T; Preg-nancy in Multiple Sclerosis Group (1998) Rate of pregnancy-related relapse in multiplesclerosis. N Engl J Med 339(5):285–291.

11. Finkelsztejn A, Brooks JBB, Paschoal FM, Jr, Fragoso YD (2011) What can we really tellwomen with multiple sclerosis regarding pregnancy? A systematic review and meta-analysis of the literature. BJOG 118(7):790–797.

12. Dendrou CA, Fugger L, Friese MA (2015) Immunopathology of multiple sclerosis. NatRev Immunol 15(9):545–558.

13. Langer-Gould A, Garren H, Slansky A, Ruiz PJ, Steinman L (2002) Late pregnancysuppresses relapses in experimental autoimmune encephalomyelitis: Evidence for asuppressive pregnancy-related serum factor. J Immunol 169(2):1084–1091.

14. Gatson NN, et al. (2011) Induction of pregnancy during established EAE halts progres-sion of CNS autoimmune injury via pregnancy-specific serum factors. J Neuroimmunol230(1–2):105–113.

15. McClain MA, et al. (2007) Pregnancy suppresses experimental autoimmune enceph-alomyelitis through immunoregulatory cytokine production. J Immunol 179(12):8146–8152.

16. Voskuhl RR, Palaszynski K (2001) Sex hormones in experimental autoimmune en-cephalomyelitis: Implications for multiple sclerosis. Neuroscientist 7(3):258–270.

17. Gold SM, Voskuhl RR (2016) Pregnancy and multiple sclerosis: From molecularmechanisms to clinical application. Semin Immunopathol 38(6):709–718.

18. Aluvihare VR, Kallikourdis M, Betz AG (2004) Regulatory T cells mediate maternaltolerance to the fetus. Nat Immunol 5(3):266–271.

19. Zenclussen AC, et al. (2006) Regulatory T cells induce a privileged tolerant microen-vironment at the fetal-maternal interface. Eur J Immunol 36(1):82–94.

20. Sakaguchi S (2005) Naturally arising Foxp3-expressing CD25+CD4+ regulatory T cellsin immunological tolerance to self and non-self. Nat Immunol 6(4):345–352.

21. Patas K, Engler JB, Friese MA, Gold SM (2013) Pregnancy and multiple sclerosis: Feto-maternal immune cross talk and its implications for disease activity. J Reprod Immunol97(1):140–146.

22. Rowe JH, Ertelt JM, Xin L, Way SS (2012) Pregnancy imprints regulatory memory thatsustains anergy to fetal antigen. Nature 490(7418):102–106.

23. Rowe JH, Ertelt JM, Aguilera MN, Farrar MA, Way SS (2011) Foxp3(+) regulatory T cellexpansion required for sustaining pregnancy compromises host defense againstprenatal bacterial pathogens. Cell Host Microbe 10(1):54–64.

24. Mao G, et al. (2010) Progesterone increases systemic and local uterine proportions ofCD4+CD25+ Treg cells during midterm pregnancy in mice. Endocrinology 151(11):5477–5488.

25. Lee JH, Ulrich B, Cho J, Park J, Kim CH (2011) Progesterone promotes differentiationof human cord blood fetal T cells into T regulatory cells but suppresses their differ-entiation into Th17 cells. J Immunol 187(4):1778–1787.

26. Wildin RS, et al. (2001) X-linked neonatal diabetes mellitus, enteropathy and endo-crinopathy syndrome is the human equivalent of mouse scurfy. Nat Genet 27(1):18–20.

27. Bennett CL, et al. (2001) The immune dysregulation, polyendocrinopathy, enteropa-thy, X-linked syndrome (IPEX) is caused by mutations of FOXP3. Nat Genet 27(1):20–21.

28. Humrich JY, et al. (2010) Homeostatic imbalance of regulatory and effector T cells dueto IL-2 deprivation amplifies murine lupus. Proc Natl Acad Sci USA 107(1):204–209.

29. Engler JB, et al. (2011) Unmasking of autoreactive CD4 T cells by depletion of CD25regulatory T cells in systemic lupus erythematosus. Ann Rheum Dis 70(12):2176–2183.

30. Nie H, et al. (2013) Phosphorylation of FOXP3 controls regulatory T cell function and isinhibited by TNF-α in rheumatoid arthritis. Nat Med 19(3):322–328.

31. Munoz-Suano A, Kallikourdis M, Sarris M, Betz AG (2012) Regulatory T cells protectfrom autoimmune arthritis during pregnancy. J Autoimmun 38(2–3):J103–J108.

32. Baca Jones C, et al. (2014) Regulatory T cells control diabetes without compromisingacute anti-viral defense. Clin Immunol 153(2):298–307.

33. Viglietta V, Baecher-Allan C, Weiner HL, Hafler DA (2004) Loss of functional sup-pression by CD4+CD25+ regulatory T cells in patients with multiple sclerosis. J ExpMed 199(7):971–979.

34. Carbone F, et al. (2014) Regulatory T cell proliferative potential is impaired in humanautoimmune disease. Nat Med 20(1):69–74.

35. Sellebjerg F, Krakauer M, Khademi M, Olsson T, Sørensen PS (2012) FOXP3, CBLB andITCH gene expression and cytotoxic T lymphocyte antigen 4 expression on CD4(+)CD25(high) T cells in multiple sclerosis. Clin Exp Immunol 170(2):149–155.

36. Huan J, et al. (2005) Decreased FOXP3 levels in multiple sclerosis patients. J NeurosciRes 81(1):45–52.

37. Venken K, et al. (2008) Compromised CD4+ CD25(high) regulatory T-cell function inpatients with relapsing-remitting multiple sclerosis is correlated with a reduced fre-quency of FOXP3-positive cells and reduced FOXP3 expression at the single-cell level.Immunology 123(1):79–89.

38. Holinka CF, Tseng YC, Finch CE (1979) Reproductive aging in C57BL/6J mice: Plasmaprogesterone, viable embryos and resorption frequency throughout pregnancy. BiolReprod 20(5):1201–1211.

39. Solano ME, et al. (2015) Progesterone and HMOX-1 promote fetal growth by CD8+T cell modulation. J Clin Invest 125(4):1726–1738.

40. Yates MA, et al. (2010) Progesterone treatment reduces disease severity and increases IL-10 in experimental autoimmune encephalomyelitis. J Neuroimmunol 220(1–2):136–139.

41. Ojasoo T, Doré JC, Gilbert J, Raynaud JP (1988) Binding of steroids to the progestinand glucocorticoid receptors analyzed by correspondence analysis. J Med Chem 31(6):1160–1169.

42. Tronche F, et al. (1999) Disruption of the glucocorticoid receptor gene in the nervoussystem results in reduced anxiety. Nat Genet 23(1):99–103.

43. Orban PC, Chui D, Marth JD (1992) Tissue- and site-specific DNA recombination intransgenic mice. Proc Natl Acad Sci USA 89(15):6861–6865.

44. Baumann S, et al. (2005) Glucocorticoids inhibit activation-induced cell death (AICD)via direct DNA-dependent repression of the CD95 ligand gene by a glucocorticoidreceptor dimer. Blood 106(2):617–625.

45. John S, et al. (2011) Chromatin accessibility pre-determines glucocorticoid receptorbinding patterns. Nat Genet 43(3):264–268.

46. van der Laan S, et al. (2008) Chromatin immunoprecipitation scanning identifiesglucocorticoid receptor binding regions in the proximal promoter of a ubiquitouslyexpressed glucocorticoid target gene in brain. J Neurochem 106(6):2515–2523.

47. Ayroldi E, et al. (2007) GILZ mediates the antiproliferative activity of glucocorticoidsby negative regulation of Ras signaling. J Clin Invest 117(6):1605–1615.

48. Niki�c I, et al. (2011) A reversible form of axon damage in experimental autoimmuneencephalomyelitis and multiple sclerosis. Nat Med 17(4):495–499.

49. Schattling B, et al. (2012) TRPM4 cation channel mediates axonal and neuronal de-generation in experimental autoimmune encephalomyelitis and multiple sclerosis.Nat Med 18(12):1805–1811.

50. Kahn DA, Baltimore D (2010) Pregnancy induces a fetal antigen-specific maternalT regulatory cell response that contributes to tolerance. Proc Natl Acad Sci USA107(20):9299–9304.

51. Robertson SA, et al. (2009) Seminal fluid drives expansion of the CD4+CD25+ T reg-ulatory cell pool and induces tolerance to paternal alloantigens in mice. Biol Reprod80(5):1036–1045.

52. Moldenhauer LM, et al. (2009) Cross-presentation of male seminal fluid antigenselicits T cell activation to initiate the female immune response to pregnancy.J Immunol 182(12):8080–8093.

53. Erlebacher A, Vencato D, Price KA, Zhang D, Glimcher LH (2007) Constraints in anti-gen presentation severely restrict T cell recognition of the allogeneic fetus. J ClinInvest 117(5):1399–1411.

54. Heng TSP, Painter MW; Immunological Genome Project Consortium (2008) The Im-munological Genome Project: Networks of gene expression in immune cells. NatImmunol 9(10):1091–1094.

55. Barlow SM, Morrison PJ, Sullivan FM (1973) Plasma corticosterone levels duringpregnancy in the mouse. Br J Pharmacol 48(2):346P.

56. Lépine S, Sulpice J-C, Giraud F (2005) Signaling pathways involved in glucocorticoid-induced apoptosis of thymocytes. Crit Rev Immunol 25(4):263–288.

57. Hoffman-Goetz L, Fietsch CL (2002) Lymphocyte apoptosis in ovariectomized micegiven progesterone and voluntary exercise. J Sports Med Phys Fitness 42(4):481–487.

58. Gold SM, et al. (2012) Dynamic development of glucocorticoid resistance during au-toimmune neuroinflammation. J Clin Endocrinol Metab 97(8):E1402–E1410.

59. Wüst S, et al. (2008) Peripheral T cells are the therapeutic targets of glucocorticoids inexperimental autoimmune encephalomyelitis. J Immunol 180(12):8434–8443.

60. Schweingruber N, et al. (2014) Chemokine-mediated redirection of T cells constitutesa critical mechanism of glucocorticoid therapy in autoimmune CNS responses. ActaNeuropathol 127(5):713–729.

61. van den Brandt J, et al. (2007) Enhanced glucocorticoid receptor signaling in T cellsimpacts thymocyte apoptosis and adaptive immune responses. Am J Pathol 170(3):1041–1053.

62. Andersen KG, Nissen JK, Betz AG (2012) Comparative genomics reveals key gain-of-function events in Foxp3 during regulatory T cell evolution. Front Immunol 3:113.

63. Sicotte NL, et al. (2002) Treatment of multiple sclerosis with the pregnancy hormoneestriol. Ann Neurol 52(4):421–428.

64. Pozzilli C, et al. (2015) Oral contraceptives combined with interferon β in multiplesclerosis. Neurol Neuroimmunol Neuroinflamm 2(4):e120.

65. Voskuhl RR, et al. (2016) Estriol combined with glatiramer acetate for women withrelapsing-remitting multiple sclerosis: A randomised, placebo-controlled, phase 2 tri-al. Lancet Neurol 15(1):35–46.

66. Lydon JP, et al. (1995) Mice lacking progesterone receptor exhibit pleiotropic re-productive abnormalities. Genes Dev 9(18):2266–2278.

E190 | www.pnas.org/cgi/doi/10.1073/pnas.1617115114 Engler et al.